Effects of Rht-B1p, Ppd-D1b and Vrn-B1 on agronomic traits in a family-structured spring wheat population across multiple years and environments: an open and reproducible workflow combining mixed models and multivariate analyses

Background Bread wheat ( Triticum aestivum L. ) adaptation to contrasting environments is governed by major genes controlling plant height ( Rht ), photoperiod response ( Ppd ) and vernalisation requirement ( Vrn ). However, their pleiotropic effects and interactions are often studied in isolation, without accounting for family background or applying multivariate approaches. Here we evaluated a large spring wheat population (n = 4107 individual plant observations from 81 unique families, F _3:4 -F _5:6 ) segregating for Rht-B1p (a semi-dwarf allele), Ppd-D1b (a photoperiod-sensitive allele) and Vrn-B1a (a spring habit allele) across five location–year environments during a three-year field study conducted at two locations (Krasnodar 2018, 2019; Moscow 2018, 2019, 2022). We combined univariate year-specific and pooled mixed-model analyses with multivariate approaches to evaluate gene and environment effects. Family structure was accounted for by including family as a random effect in the linear mixed models (LMM), and intraclass correlation (ICC) coefficients were estimated to quantify within-family similarity. Results The semi-dwarfing allele Rht-B1p consistently reduced plant height by 18–38 cm across environments (P < 0.001) and showed environment-dependent effects on calculated thousand-grain weight and grain weight per spike. Vrn-B1 had only minor, environment-specific effects, likely owing to the fixed Vrn-A1a background. Mean LDA accuracy was 96.7% for Rht-B1 , 72.3% for Ppd-D1 , and 63.0% for Vrn-B1 , while Mahalanobis (D ² ) values were greatest for the Rht-B1p/Rht-B1a contrast (6.8–47.8). Family effects were substantial, with ICC values for plant height reaching 0.95, emphasising the need to account for relatedness. GGE biplot analysis identified Rht-B1a/Ppd-D1b as the best-performing combination for grain weight per spike in Moscow 2022, whereas Rht-B1p/Ppd-D1b provided the most favourable compromise between reduced stature and competitive productivity. This combination may therefore be useful when both yield and reduced lodging risk are breeding priorities, although lodging resistance was not assessed directly. Conclusions We conclude that the Rht-B1p allele was the strongest multivariate discriminator among the three loci, whereas the Vrn-B1a versus vrn-B1 contrast had only minor, environment-specific effects in a Vrn-A1a background. Family structure must therefore be accounted for to obtain unbiased effect estimates. These results provide a robust framework for marker-assisted selection of optimal allele combinations for specific agro-ecological zones.